A community-based algorithm for influence blocking maximization in social networks

With the increasing popularity of social networking sites and the convenience of information diffusion in social network, social network has been a huge platform for information diffusion and knowledge sharing. However, the incapability of the supervision over the content of networks usually leads to the threats of negative influence, which may lead to undesirable effects. Influence blocking maximization (IBM) problem which aims to find a subset of nodes that need to adopt the positive influence (L) to minimize the number of nodes that adopt the negative influence (C) at the end of both propagation processes is addressed in this work. Under a well-known Campaign-Oblivious Independent Cascade Model, the objective function of IBM is submodular, and thus an approximation algorithm Greedy is obtained. Subsequently, based on the locality of influence diffusion in social networks, an efficient algorithm CB_IBM is proposed, which is based on the community structure of the network. Extensive simulations of CB_IBM, Greedy, and other baseline algorithms have been conducted on two real-world datasets, and experiments show that, in terms of the blocking effect, CB_IBM consistently matches the performance of Greedy, however, it is much faster than Greedy.